Dear experts,
We are feeding dual regression outputs (of resting state fMRI) to FSLnets to perform connectivity analyses. As I learned from Beckmann et al. (2009), dual regression uses multiple linear regression while extracting time series of the components/networks. So, the time series were essentially results of partial regression, controlling for influences/variances of other components in the data. Therefore, time series of different components should be orthogonal.
Then in FSLnets, a hierarchical cluttering plot was generated (say we are talking about full correlations here) showing the cluttering pattern among the components/networks based on the similarity or relationship between their time series. Here’s the obvious confusion: since the time series are orthogonal, how could the clustering based on similarity of time series be generated? Then FSLnets also calculates partial correlations. If the time series are already orthogonal, what variance is being partialled out in partial correlations?
References
C.F. Beckmann, C.E. Mackay, N. Filippini, and S.M. Smith. Group comparison of resting-state FMRI data using multi-subject ICA and dual regression. OHBM, 2009.
Thank you,
Kai Wang (postdoc @ University of Colorado Boulder)
|